ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.pod
Revision: 1.28
Committed: Fri Jun 10 11:51:03 2011 UTC (13 years ago) by root
Branch: MAIN
Changes since 1.27: +1 -1 lines
Log Message:
*** empty log message ***

File Contents

# User Rev Content
1 root 1.14 =head1 LIBECB - e-C-Builtins
2 root 1.3
3 root 1.14 =head2 ABOUT LIBECB
4    
5     Libecb is currently a simple header file that doesn't require any
6     configuration to use or include in your project.
7    
8 sf-exg 1.16 It's part of the e-suite of libraries, other members of which include
9 root 1.14 libev and libeio.
10    
11     Its homepage can be found here:
12    
13     http://software.schmorp.de/pkg/libecb
14    
15     It mainly provides a number of wrappers around GCC built-ins, together
16     with replacement functions for other compilers. In addition to this,
17 sf-exg 1.16 it provides a number of other lowlevel C utilities, such as endianness
18 root 1.14 detection, byte swapping or bit rotations.
19    
20 root 1.24 Or in other words, things that should be built into any standard C system,
21     but aren't, implemented as efficient as possible with GCC, and still
22     correct with other compilers.
23 root 1.17
24 root 1.14 More might come.
25 root 1.3
26     =head2 ABOUT THE HEADER
27    
28 root 1.14 At the moment, all you have to do is copy F<ecb.h> somewhere where your
29     compiler can find it and include it:
30    
31     #include <ecb.h>
32    
33     The header should work fine for both C and C++ compilation, and gives you
34     all of F<inttypes.h> in addition to the ECB symbols.
35    
36 sf-exg 1.16 There are currently no object files to link to - future versions might
37 root 1.14 come with an (optional) object code library to link against, to reduce
38     code size or gain access to additional features.
39    
40     It also currently includes everything from F<inttypes.h>.
41    
42     =head2 ABOUT THIS MANUAL / CONVENTIONS
43    
44     This manual mainly describes each (public) function available after
45     including the F<ecb.h> header. The header might define other symbols than
46     these, but these are not part of the public API, and not supported in any
47     way.
48    
49     When the manual mentions a "function" then this could be defined either as
50     as inline function, a macro, or an external symbol.
51    
52     When functions use a concrete standard type, such as C<int> or
53     C<uint32_t>, then the corresponding function works only with that type. If
54     only a generic name is used (C<expr>, C<cond>, C<value> and so on), then
55     the corresponding function relies on C to implement the correct types, and
56     is usually implemented as a macro. Specifically, a "bool" in this manual
57     refers to any kind of boolean value, not a specific type.
58 root 1.1
59     =head2 GCC ATTRIBUTES
60    
61 root 1.20 A major part of libecb deals with GCC attributes. These are additional
62 sf-exg 1.26 attributes that you can assign to functions, variables and sometimes even
63 root 1.20 types - much like C<const> or C<volatile> in C.
64    
65     While GCC allows declarations to show up in many surprising places,
66 sf-exg 1.26 but not in many expected places, the safest way is to put attribute
67 root 1.20 declarations before the whole declaration:
68    
69     ecb_const int mysqrt (int a);
70     ecb_unused int i;
71    
72     For variables, it is often nicer to put the attribute after the name, and
73     avoid multiple declarations using commas:
74    
75     int i ecb_unused;
76 root 1.3
77 root 1.1 =over 4
78    
79 root 1.2 =item ecb_attribute ((attrs...))
80 root 1.1
81 root 1.15 A simple wrapper that expands to C<__attribute__((attrs))> on GCC, and to
82     nothing on other compilers, so the effect is that only GCC sees these.
83    
84     Example: use the C<deprecated> attribute on a function.
85    
86     ecb_attribute((__deprecated__)) void
87     do_not_use_me_anymore (void);
88 root 1.2
89 root 1.3 =item ecb_unused
90    
91     Marks a function or a variable as "unused", which simply suppresses a
92     warning by GCC when it detects it as unused. This is useful when you e.g.
93     declare a variable but do not always use it:
94    
95 root 1.15 {
96     int var ecb_unused;
97 root 1.3
98 root 1.15 #ifdef SOMECONDITION
99     var = ...;
100     return var;
101     #else
102     return 0;
103     #endif
104     }
105 root 1.3
106 root 1.2 =item ecb_noinline
107    
108 root 1.9 Prevent a function from being inlined - it might be optimised away, but
109 root 1.3 not inlined into other functions. This is useful if you know your function
110     is rarely called and large enough for inlining not to be helpful.
111    
112 root 1.2 =item ecb_noreturn
113    
114 root 1.17 Marks a function as "not returning, ever". Some typical functions that
115     don't return are C<exit> or C<abort> (which really works hard to not
116     return), and now you can make your own:
117    
118     ecb_noreturn void
119     my_abort (const char *errline)
120     {
121     puts (errline);
122     abort ();
123     }
124    
125 sf-exg 1.19 In this case, the compiler would probably be smart enough to deduce it on
126     its own, so this is mainly useful for declarations.
127 root 1.17
128 root 1.2 =item ecb_const
129    
130 sf-exg 1.19 Declares that the function only depends on the values of its arguments,
131 root 1.17 much like a mathematical function. It specifically does not read or write
132     any memory any arguments might point to, global variables, or call any
133     non-const functions. It also must not have any side effects.
134    
135     Such a function can be optimised much more aggressively by the compiler -
136     for example, multiple calls with the same arguments can be optimised into
137     a single call, which wouldn't be possible if the compiler would have to
138     expect any side effects.
139    
140     It is best suited for functions in the sense of mathematical functions,
141 sf-exg 1.19 such as a function returning the square root of its input argument.
142 root 1.17
143     Not suited would be a function that calculates the hash of some memory
144     area you pass in, prints some messages or looks at a global variable to
145     decide on rounding.
146    
147     See C<ecb_pure> for a slightly less restrictive class of functions.
148    
149 root 1.2 =item ecb_pure
150    
151 root 1.17 Similar to C<ecb_const>, declares a function that has no side
152     effects. Unlike C<ecb_const>, the function is allowed to examine global
153     variables and any other memory areas (such as the ones passed to it via
154     pointers).
155    
156     While these functions cannot be optimised as aggressively as C<ecb_const>
157     functions, they can still be optimised away in many occasions, and the
158     compiler has more freedom in moving calls to them around.
159    
160     Typical examples for such functions would be C<strlen> or C<memcmp>. A
161     function that calculates the MD5 sum of some input and updates some MD5
162     state passed as argument would I<NOT> be pure, however, as it would modify
163     some memory area that is not the return value.
164    
165 root 1.2 =item ecb_hot
166    
167 root 1.17 This declares a function as "hot" with regards to the cache - the function
168     is used so often, that it is very beneficial to keep it in the cache if
169     possible.
170    
171     The compiler reacts by trying to place hot functions near to each other in
172     memory.
173    
174 sf-exg 1.19 Whether a function is hot or not often depends on the whole program,
175 root 1.17 and less on the function itself. C<ecb_cold> is likely more useful in
176     practise.
177    
178 root 1.2 =item ecb_cold
179    
180 root 1.17 The opposite of C<ecb_hot> - declares a function as "cold" with regards to
181     the cache, or in other words, this function is not called often, or not at
182     speed-critical times, and keeping it in the cache might be a waste of said
183     cache.
184    
185     In addition to placing cold functions together (or at least away from hot
186     functions), this knowledge can be used in other ways, for example, the
187     function will be optimised for size, as opposed to speed, and codepaths
188     leading to calls to those functions can automatically be marked as if
189 root 1.27 C<ecb_expect_false> had been used to reach them.
190 root 1.17
191     Good examples for such functions would be error reporting functions, or
192     functions only called in exceptional or rare cases.
193    
194 root 1.2 =item ecb_artificial
195    
196 root 1.17 Declares the function as "artificial", in this case meaning that this
197     function is not really mean to be a function, but more like an accessor
198     - many methods in C++ classes are mere accessor functions, and having a
199     crash reported in such a method, or single-stepping through them, is not
200     usually so helpful, especially when it's inlined to just a few instructions.
201    
202     Marking them as artificial will instruct the debugger about just this,
203     leading to happier debugging and thus happier lives.
204    
205     Example: in some kind of smart-pointer class, mark the pointer accessor as
206     artificial, so that the whole class acts more like a pointer and less like
207     some C++ abstraction monster.
208    
209     template<typename T>
210     struct my_smart_ptr
211     {
212     T *value;
213    
214     ecb_artificial
215     operator T *()
216     {
217     return value;
218     }
219     };
220    
221 root 1.2 =back
222 root 1.1
223     =head2 OPTIMISATION HINTS
224    
225     =over 4
226    
227 root 1.14 =item bool ecb_is_constant(expr)
228 root 1.1
229 root 1.3 Returns true iff the expression can be deduced to be a compile-time
230     constant, and false otherwise.
231    
232     For example, when you have a C<rndm16> function that returns a 16 bit
233     random number, and you have a function that maps this to a range from
234 root 1.5 0..n-1, then you could use this inline function in a header file:
235 root 1.3
236     ecb_inline uint32_t
237     rndm (uint32_t n)
238     {
239 root 1.6 return (n * (uint32_t)rndm16 ()) >> 16;
240 root 1.3 }
241    
242     However, for powers of two, you could use a normal mask, but that is only
243     worth it if, at compile time, you can detect this case. This is the case
244     when the passed number is a constant and also a power of two (C<n & (n -
245     1) == 0>):
246    
247     ecb_inline uint32_t
248     rndm (uint32_t n)
249     {
250     return is_constant (n) && !(n & (n - 1))
251     ? rndm16 () & (num - 1)
252 root 1.6 : (n * (uint32_t)rndm16 ()) >> 16;
253 root 1.3 }
254    
255 root 1.14 =item bool ecb_expect (expr, value)
256 root 1.1
257 root 1.7 Evaluates C<expr> and returns it. In addition, it tells the compiler that
258     the C<expr> evaluates to C<value> a lot, which can be used for static
259     branch optimisations.
260 root 1.1
261 root 1.27 Usually, you want to use the more intuitive C<ecb_expect_true> and
262     C<ecb_expect_false> functions instead.
263 root 1.1
264 root 1.27 =item bool ecb_expect_true (cond)
265 root 1.1
266 root 1.27 =item bool ecb_expect_false (cond)
267 root 1.1
268 root 1.7 These two functions expect a expression that is true or false and return
269     C<1> or C<0>, respectively, so when used in the condition of an C<if> or
270     other conditional statement, it will not change the program:
271    
272     /* these two do the same thing */
273     if (some_condition) ...;
274 root 1.27 if (ecb_expect_true (some_condition)) ...;
275 root 1.7
276 root 1.27 However, by using C<ecb_expect_true>, you tell the compiler that the
277     condition is likely to be true (and for C<ecb_expect_false>, that it is
278     unlikely to be true).
279 root 1.7
280 root 1.9 For example, when you check for a null pointer and expect this to be a
281 root 1.27 rare, exceptional, case, then use C<ecb_expect_false>:
282 root 1.7
283     void my_free (void *ptr)
284     {
285 root 1.27 if (ecb_expect_false (ptr == 0))
286 root 1.7 return;
287     }
288    
289     Consequent use of these functions to mark away exceptional cases or to
290     tell the compiler what the hot path through a function is can increase
291     performance considerably.
292    
293 root 1.27 You might know these functions under the name C<likely> and C<unlikely>
294     - while these are common aliases, we find that the expect name is easier
295     to understand when quickly skimming code. If you wish, you can use
296     C<ecb_likely> instead of C<ecb_expect_true> and C<ecb_unlikely> instead of
297     C<ecb_expect_false> - these are simply aliases.
298    
299 root 1.7 A very good example is in a function that reserves more space for some
300     memory block (for example, inside an implementation of a string stream) -
301 root 1.9 each time something is added, you have to check for a buffer overrun, but
302 root 1.7 you expect that most checks will turn out to be false:
303    
304     /* make sure we have "size" extra room in our buffer */
305     ecb_inline void
306     reserve (int size)
307     {
308 root 1.27 if (ecb_expect_false (current + size > end))
309 root 1.7 real_reserve_method (size); /* presumably noinline */
310     }
311    
312 root 1.14 =item bool ecb_assume (cond)
313 root 1.7
314     Try to tell the compiler that some condition is true, even if it's not
315     obvious.
316    
317     This can be used to teach the compiler about invariants or other
318     conditions that might improve code generation, but which are impossible to
319     deduce form the code itself.
320    
321 root 1.27 For example, the example reservation function from the C<ecb_expect_false>
322 root 1.7 description could be written thus (only C<ecb_assume> was added):
323    
324     ecb_inline void
325     reserve (int size)
326     {
327 root 1.27 if (ecb_expect_false (current + size > end))
328 root 1.7 real_reserve_method (size); /* presumably noinline */
329    
330     ecb_assume (current + size <= end);
331     }
332    
333     If you then call this function twice, like this:
334    
335     reserve (10);
336     reserve (1);
337    
338     Then the compiler I<might> be able to optimise out the second call
339     completely, as it knows that C<< current + 1 > end >> is false and the
340     call will never be executed.
341    
342     =item bool ecb_unreachable ()
343    
344     This function does nothing itself, except tell the compiler that it will
345 root 1.9 never be executed. Apart from suppressing a warning in some cases, this
346 root 1.7 function can be used to implement C<ecb_assume> or similar functions.
347    
348 root 1.14 =item bool ecb_prefetch (addr, rw, locality)
349 root 1.7
350     Tells the compiler to try to prefetch memory at the given C<addr>ess
351 root 1.10 for either reading (C<rw> = 0) or writing (C<rw> = 1). A C<locality> of
352 root 1.7 C<0> means that there will only be one access later, C<3> means that
353     the data will likely be accessed very often, and values in between mean
354     something... in between. The memory pointed to by the address does not
355     need to be accessible (it could be a null pointer for example), but C<rw>
356     and C<locality> must be compile-time constants.
357    
358     An obvious way to use this is to prefetch some data far away, in a big
359 root 1.9 array you loop over. This prefetches memory some 128 array elements later,
360 root 1.7 in the hope that it will be ready when the CPU arrives at that location.
361    
362     int sum = 0;
363    
364     for (i = 0; i < N; ++i)
365     {
366     sum += arr [i]
367     ecb_prefetch (arr + i + 128, 0, 0);
368     }
369    
370     It's hard to predict how far to prefetch, and most CPUs that can prefetch
371     are often good enough to predict this kind of behaviour themselves. It
372     gets more interesting with linked lists, especially when you do some fair
373     processing on each list element:
374    
375     for (node *n = start; n; n = n->next)
376     {
377     ecb_prefetch (n->next, 0, 0);
378     ... do medium amount of work with *n
379     }
380    
381     After processing the node, (part of) the next node might already be in
382     cache.
383 root 1.1
384 root 1.2 =back
385 root 1.1
386     =head2 BIT FIDDLING / BITSTUFFS
387    
388 root 1.4 =over 4
389    
390 root 1.3 =item bool ecb_big_endian ()
391    
392     =item bool ecb_little_endian ()
393    
394 sf-exg 1.11 These two functions return true if the byte order is big endian
395     (most-significant byte first) or little endian (least-significant byte
396     first) respectively.
397    
398 root 1.24 On systems that are neither, their return values are unspecified.
399    
400 root 1.3 =item int ecb_ctz32 (uint32_t x)
401    
402 sf-exg 1.11 Returns the index of the least significant bit set in C<x> (or
403 root 1.24 equivalently the number of bits set to 0 before the least significant bit
404     set), starting from 0. If C<x> is 0 the result is undefined. A common use
405     case is to compute the integer binary logarithm, i.e., C<floor (log2
406     (n))>. For example:
407 sf-exg 1.11
408 root 1.15 ecb_ctz32 (3) = 0
409     ecb_ctz32 (6) = 1
410 sf-exg 1.11
411 root 1.3 =item int ecb_popcount32 (uint32_t x)
412    
413 sf-exg 1.11 Returns the number of bits set to 1 in C<x>. For example:
414    
415 root 1.15 ecb_popcount32 (7) = 3
416     ecb_popcount32 (255) = 8
417 sf-exg 1.11
418 root 1.8 =item uint32_t ecb_bswap16 (uint32_t x)
419    
420 root 1.3 =item uint32_t ecb_bswap32 (uint32_t x)
421    
422 root 1.21 These two functions return the value of the 16-bit (32-bit) value C<x>
423     after reversing the order of bytes (0x11223344 becomes 0x44332211).
424 sf-exg 1.13
425 root 1.3 =item uint32_t ecb_rotr32 (uint32_t x, unsigned int count)
426    
427     =item uint32_t ecb_rotl32 (uint32_t x, unsigned int count)
428    
429 root 1.22 These two functions return the value of C<x> after rotating all the bits
430 sf-exg 1.11 by C<count> positions to the right or left respectively.
431    
432 root 1.20 Current GCC versions understand these functions and usually compile them
433     to "optimal" code (e.g. a single C<roll> on x86).
434    
435 root 1.3 =back
436 root 1.1
437     =head2 ARITHMETIC
438    
439 root 1.3 =over 4
440    
441 root 1.14 =item x = ecb_mod (m, n)
442 root 1.3
443 root 1.25 Returns C<m> modulo C<n>, which is the same as the positive remainder
444     of the division operation between C<m> and C<n>, using floored
445     division. Unlike the C remainder operator C<%>, this function ensures that
446     the return value is always positive and that the two numbers I<m> and
447     I<m' = m + i * n> result in the same value modulo I<n> - in other words,
448     C<ecb_mod> implements the mathematical modulo operation, which is missing
449     in the language.
450 root 1.14
451 sf-exg 1.23 C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be
452 root 1.14 negatable, that is, both C<m> and C<-m> must be representable in its
453 root 1.25 type (this typically includes the minimum signed integer value, the same
454     limitation as for C</> and C<%> in C).
455 sf-exg 1.11
456 root 1.24 Current GCC versions compile this into an efficient branchless sequence on
457 root 1.28 almost all CPUs.
458 root 1.24
459     For example, when you want to rotate forward through the members of an
460     array for increasing C<m> (which might be negative), then you should use
461     C<ecb_mod>, as the C<%> operator might give either negative results, or
462     change direction for negative values:
463    
464     for (m = -100; m <= 100; ++m)
465     int elem = myarray [ecb_mod (m, ecb_array_length (myarray))];
466    
467 root 1.3 =back
468 root 1.1
469     =head2 UTILITY
470    
471 root 1.3 =over 4
472    
473 sf-exg 1.23 =item element_count = ecb_array_length (name)
474 root 1.3
475 sf-exg 1.13 Returns the number of elements in the array C<name>. For example:
476    
477     int primes[] = { 2, 3, 5, 7, 11 };
478     int sum = 0;
479    
480     for (i = 0; i < ecb_array_length (primes); i++)
481     sum += primes [i];
482    
483 root 1.3 =back
484 root 1.1
485